fix(api): skip bootstrap ddl statement timeout

apps/api/src/db/base.py

@@ -216,6 +216,29 @@ def _is_database_connection_budget_error(exc: BaseException) -> bool:
     return False
 
 
+def _is_database_bootstrap_ddl_timeout(exc: BaseException) -> bool:
+    """Return True only for optional bootstrap DDL canceled by DB timeout."""
+    seen: set[int] = set()
+    current: BaseException | None = exc
+    timeout_markers = (
+        "querycancelederror",
+        "canceling statement due to statement timeout",
+        "statement timeout",
+    )
+
+    while current is not None and id(current) not in seen:
+        seen.add(id(current))
+        message = f"{type(current).__name__}: {current}".lower()
+        if any(marker in message for marker in timeout_markers):
+            return True
+        current = (
+            getattr(current, "orig", None)
+            or getattr(current, "__cause__", None)
+            or getattr(current, "__context__", None)
+        )
+    return False
+
+
 async def init_db() -> None:
     """
     Initialize database tables
@@ -251,6 +274,14 @@ async def init_db() -> None:
                         timeout_seconds=_DB_BOOTSTRAP_DDL_WAIT_SECONDS,
                         lock_name=_DB_BOOTSTRAP_LOCK_NAME,
                     )
+                except DBAPIError as exc:
+                    if not _is_database_bootstrap_ddl_timeout(exc):
+                        raise
+                    logger.warning(
+                        "database_bootstrap_statement_timeout_skipped",
+                        error_type=type(exc).__name__,
+                        lock_name=_DB_BOOTSTRAP_LOCK_NAME,
+                    )
             finally:
                 await lock_conn.execute(
                     text("SELECT pg_advisory_unlock(hashtext(:lock_name))"),

apps/api/tests/test_runtime_bootstrap_guards.py

@@ -12,6 +12,7 @@ from collections.abc import Awaitable
 from typing import Any
 
 import pytest
+from sqlalchemy.exc import DBAPIError
 
 
 class _FakeScalarResult:
@@ -190,6 +191,29 @@ async def test_init_db_releases_bootstrap_lock_when_ddl_times_out(monkeypatch):
     assert any("pg_advisory_unlock" in stmt for stmt in fake_engine.lock_conn.statements)
 
 
+@pytest.mark.asyncio
+async def test_init_db_skips_bootstrap_when_postgres_statement_timeout(monkeypatch):
+    from src.db import base as db_base
+
+    fake_engine = _FakeEngine()
+
+    async def fake_run_init_db_ddl(_engine: object) -> None:
+        raise DBAPIError(
+            "ALTER TABLE approval_records ADD COLUMN IF NOT EXISTS telegram_message_id",
+            {},
+            Exception("canceling statement due to statement timeout"),
+        )
+
+    monkeypatch.setattr(db_base, "get_engine", lambda: fake_engine)
+    monkeypatch.setattr(db_base, "_run_init_db_ddl", fake_run_init_db_ddl)
+
+    await db_base.init_db()
+
+    assert "pg_try_advisory_lock" in fake_engine.lock_conn.statements[0]
+    assert any("pg_advisory_unlock" in stmt for stmt in fake_engine.lock_conn.statements)
+    assert "COMMIT" in fake_engine.lock_conn.statements
+
+
 @pytest.mark.asyncio
 async def test_signal_worker_initializes_worker_redis_pool_before_tasks(monkeypatch):
     from src.workers import signal_worker

docs/LOGBOOK.md

@@ -1,3 +1,18 @@
+## 2026-07-02 — 14:12 API rollout CrashLoop root cause 與 bootstrap DDL timeout 修復
+
+**完成內容**：
+- `f9469bcc2` CD rollout 期間 production API 變成 `502`；read-only K8s 查詢確認 `awoooi-api` 新 pod `CrashLoopBackOff`，web / worker / auto-repair-canary 皆 Running。
+- `kubectl logs --previous` 顯示 API startup 在 `init_db()` 執行 `ALTER TABLE approval_records ADD COLUMN IF NOT EXISTS telegram_message_id...` 時被 PostgreSQL `statement_timeout` 取消，SQLAlchemy 包成 `DBAPIError` 後重新 raise，導致 container exit 3。
+- `apps/api/src/db/base.py` 新增 `_is_database_bootstrap_ddl_timeout()`，只針對 optional bootstrap DDL 的 `QueryCanceledError` / `canceling statement due to statement timeout` fail-visible skip；其他 DBAPIError 仍 raise，避免吞真正 migration / SQL 錯誤。
+- `apps/api/tests/test_runtime_bootstrap_guards.py` 新增 regression：模擬 `DBAPIError(... statement timeout ...)` 時，`init_db()` 必須釋放 advisory lock、commit unlock，且不讓 API startup crash。
+
+**驗證**：
+- `python3.11 -m py_compile apps/api/src/db/base.py`
+- `DATABASE_URL=postgresql+asyncpg://test:test@localhost/test python3.11 -m pytest apps/api/tests/test_runtime_bootstrap_guards.py -q`：`9 passed`。
+
+**仍維持**：
+- 未使用 GitHub / `gh` / GitHub API；未讀 secret / token / `.env` / raw sessions / SQLite / auth；未重啟主機 / Docker / Nginx / K3s / DB / firewall；K8s 查詢為 read-only。
+
 ## 2026-07-02 — 13:55 Telegram 告警 receipt / AI route coverage 缺口讀回
 
 **完成內容**：